Another example is that if we want to conduct a survey where our target population is all the poly students in Singapore, where we wanted to find out how the workload affects their mental health, if we made use of convenience sampling - one way we would gather information could be by standing outside at the entrance of RP to gather answers from students. However, this would then result in an inaccurate representation of the rest of the population, as we would only be getting responses from the students in one polytechnic. when we should be gathering responses from all 5 polytechnics in Singapore, and not just RP. 

Due to the fact that statistics can be manipulated with numbers, this example also shows the importance of choosing a random or non-random sampling method. Where a non-random method would better suit a survey on a specific population. And a random sampling method would be used to study a whole population. In the case of the example shown, we should use purposive sampling where we have an equal number of students drawn from each polytechnic in Singapore, and also from each school. Having an equal number of students drawn from each polytechnic and from each year of study (Y1, Y2, Y3) would ensure that the p-value is lowered and the population is accurately represented.  

 A sample size of 25 would reduce the chances of “randomness” in our answers, lowering the p-value of the data. P-value refers to the confidence level of the data. A lower p-value would mean there is little to no errors in the data. 

 With a larger sample size, we are able to spot the common characteristics of our data as compared to a smaller size where the opinions are limited. 

The statistical significance in data is calculated using a p-value. If the sample design does not represent the population of our research, there is a likelihood of the results having inaccuracies/errors.

The lower the p-value, the lower the chance our data would be due to chance/error. Meaning that our data is probably correct. 

 

If the sample design is prone to giving errors in data, the p-value would be high. The higher the p-value, the more likely it will be for the sample design to bring about errors and inaccuracies in the data.

 

Through the p-value, we are able to tell whether or not the sampling structure we have chosen is the most suitable.

 

An example of a good sampling structure is one that we can use to obtain the most precise data through a random sampling method. A random sampling method would ensure that there is an accurate representation of the population that we have chosen. Unlike, non-random sampling, where there is a higher possibility of an ‘error’, due to the fact that the chosen sample size may be selected from biasness or is an inaccurate selection of the sample population.  Another advantage of random sampling is that the sample will represent the target population and eliminate sampling bias. It would also ensure that everyone is given an equal chance to be selected. 

 

From this, we derived that we would use the stratified sampling method. Stratified random sampling involves the division of a population into several homogeneous subgroups, then using simple random sampling in each subgroup. It has more statistical precision than simple random sampling. This is true if the strata/groups are homogenous. It also assures that you can represent not just the overall population, but also key sub-groups of the population, especially the minority sub groups.